Serum CXCL13 positively correlates with prostatic disease, prostate-specific antigen and mediates prostate cancer cell invasion, integrin clustering and cell adhesion

Abstract

Chemokines and corresponding receptor interactions have been shown to be involved in prostate cancer (PCa) progression and organ-specific metastasis. We have recently shown that PCa cell lines and primary prostate tumors express CXCR5, which correlates with PCa grade. In this study, we present the first evidence that CXCL13, the only ligand for CXCR5, and IL-6 were significantly elevated in PCa patient serum compared to serum from subjects with benign prostatic hyperplasia (BPH), or high-grade prostatic intraepithelial neoplasia (HGPIN) as well as normal healthy donors (NHD). Serum CXCL13 levels significantly (p < 0.0001) correlated with serum prostate-specific antigen (PSA), whereas serum IL-6 levels significantly (p < 0.0003) correlated with CXCL13 serum levels. CXCL13 was found to be a better predictor of PCa than PSA. In addition, CXCL13 was highly expressed by human bone marrow endothelial (HBME) cells and osteoblasts (OBs), but not osteoclasts (OCs), following treatment with physiologically relevant levels of interleukin-6 (IL-6). We further demonstrate that CXCL13, produced by IL-6-treated HBME cells, was able to induce PCa cell invasion in a CXCR5-dependent manner. CXCL13-mediated PCa cell α_vβ₃-integrin clustering and adhesion to HBME cells was abrogated by CXCR5 blockade. These results demonstrate that the CXCL13-CXCR5 axis is significantly associated with PCa progression.

1. Introduction

PCa is one of the leading causes of cancer-related deaths among men in US [1]. Despite recent advancements in early diagnosis (e.g., PSA monitoring and digital rectal examination) and treatment, metastasis is still a major cause of mortality in PCa [2]. Unfortunately, the mechanisms involved in the malignant process are not completely understood [3; 4] and PSA can not be used to solely predict the extent of prostatic disease [5]. Introduction of PSA as a screening tool has had a dual effect – dramatically decreasing the incidence of advanced PCa and increasing the number of new cases of PCa [6]. However, men with normal levels of total serum PSA can have PCa. Indeed, 15% of PCa cases with positive biopsies had PSA levels < 4 ng/mL and a weak association of PSA with tumor volume [5; 7]. Taken together, these studies demonstrate the great need for additional biomarkers to improve or complement the use of PSA that might also correlate with advanced disease.

Bone metastases are frequent complications of many cancers and occur in up to 70% of patients with advanced PCa [8]. The exact incidence of bone metastasis is unknown, but it is estimated that 350,000 people diagnosed with cancer die with bone (osteolytic or osteoblastic) metastases annually in the US [9]. PCa lesions are predominantly osteoblastic [10]. Tumor cells produce adhesive molecules that allow them to bind bone marrow (i.e., stromal cells, bone matrix proteins, etc.). These adhesive interactions cause tumor cells to colonize in this environment as well as increase the production of angiogenic and bone-resorbing factors for enhanced tumor growth. In this regard, studies from our laboratory and others indicate that multiple chemokines and their corresponding receptors (including CCR5, CCR7, CCR9, CXCR3, CXCR4, CXCR5, and CXCR7) may be involved in the multi-step and dynamic process of PCa cell metastasis [11; 12; 13; 14; 15; 16; 17; 18; 19; 20; 21; 22]. In this study, we show that CXCL13 and IL-6 is highly elevated in PCa patient sera and positively correlated with serum PSA. Surprisingly, serum CXCL13 levels seem to be a better predictor of advanced disease than PSA. We also show that HBME cells and OBs secrete CXCL13 after IL-6 stimulation, which is known to be highly elevated in the sera of patients with metastatic disease [23]. Lastly, we demonstrate that the CXCR5 axis is involved in α_vβ₃ integrin clustering for adhesion of PCa cells to HBME cells.

2. Material and methods

2.1 Cell lines and cell culture

PC3 cell lines were obtained from ATCC, Dr. Carlton Cooper (University of Delaware) provided HBME cells, human OB and OC cell lines were obtained from Clonetics. PC3 cell lines were cultured in Ham’s F12K medium with 2mM L-glutamine and adjusted to contain 1.5 g/L sodium bicarbonate (ATCC) with 10% fetal bovine serum (FBS) (Sigma). After five passages in Ham’s F12K media, PC3 cells were switched to RPMI-1640 with 10% fetal bovine serum FBS. HBME cells were cultured in Dulbecco's Modified Eagle Medium (DMED) supplemented with penicillin, streptomycin and 10% FBS. Human OB and OC cells were cultured in corresponding select medium (Clonetics) along with with OB or OC growth factors (Clonetics), respectively, supplemented with gentamycin, amphotericin-B and 10% FBS according to manufacturer’s instructions. All cell lines were cultured at 37°C with 5% CO₂.

2.2 CXCL13, IL-1β, IL-6, and TNF-α ELISAs

Serum from a cohort of untreated patients diagnosed with PCa (n= 15), BPH (n = 10), or HGPIN (n = 5), and healthy donors (n = 10) were obtained from the Cooperative Cancer Tissue Resource Center at the University of Alabama at Birmingham. Healthy donors had no active urologic disease or symptoms at the time of blood collection. All subjects gave written informed consent and were approved by the University of Alabama at Birmingham Institutional Review Board (IRB). Subsequently, the University of Louisville IRB approved the use of these diagnostic specimens in accordance with the Department of Health and Human Service Policy for the Protection of Human Research Subjects 45 CFR 46.101(b) 2 and use of archived de-identified materials.

ELISA kits for total serum PSA was purchased from ALPCO Diagnostics, human CXCL13, IL-1β, IL-6, and TNF-α ELISA kits were obtained from R&D Systems and completed according to manufacturers’ instructions. HBME cells, OBs and OCs were seeded in 48 well plates (5 × 10⁴ cells/well). Conditioned medium form untreated cells and cells treated with IL-6 (17.5 pg/mL) were collected after incubation overnight (18 hours). ELISA assays were performed to quantify CXCL13 in conditioned medium collected form the untreated and treated cells according to manufacturers protocols. ELISA assays were capable of detecting > 1 pg/mL of PSA, CXCL13, IL-1β, IL-6, or TNF-α.

2.3 Invasion assay

PCa cell invasion toward HBME cells, OC or OB cultures was measured using BD Bio-coat Fluroblock tumor invasion chambers (BD Bioscience) according to manufacturer’s protocols with minor modifications. Briefly, HBME, OC, or OB cells (5 × 10⁴ cells/well) were plated in 24 well companion (bottom chamber) plates in triplicate and cultured with 5% CO₂ at 37°C for 16 hours and treated with physiologically relevant levels of recombinant IL-6 (R&D Systems) that are found in normal disease-free human serum (1.0 pg/mL of IL-6), patients with localized PCa (1.5 pg/mL of IL-6), or patients with metastatic PCa (17.5 pg/mL of IL-6) [24; 25; 26; 27]. Untreated cells were used as control. DMEM was added to top chamber of Fluroblock tumor invasion chambers and incubated for 2 hours with 5% CO₂ at 37°C. Invasion chambers were next placed on top of HBME, OC, or OB cell cultures growing in bottom chambers. PC3 cells (10⁵ cells in 500 µL) along with 1.0 µg/mL of anti-CXCR5 or isotype control antibodies (Abs) (R&D system) were added to the top chamber, and allowed to invade the chamber matrix during incubation with 5% CO₂ at 37°C for 20 hours. Next, invasion chamber inserts were transferred to another 24 well plate containing 4 µg/mL calcein-AM (Invitrogen) in 500 µl of Hank’s buffered salt solution (HBSS) and incubated with 5% CO₂ at 37°C for 30 minutes for post invasion labeling. Unbound calcein-AM was removed after 2 exchanges in HBSS and plates were read using a fluorescence reader at excitation 485 nm and emission 530 nm.

2.4 Cell adhesion assay

HBME, OC and OB cells (5 × 10⁴ cell/well) were treated for 18 hours as before with physiologically relevant levels of recombinant IL-6. Next, untreated or PC3 cells (10⁵ cells in 500 µL) were treated with 1.0 µg/mL of anti-CXCR5 Ab for one hour and incubated with 1 mL of calcein-AM solution (4 µg/mL in HBSS) for 30 minutes with 5% CO₂ at 37°C. Cells were washed with serum free RPMI medium to remove unbound calcein-AM. Next, PC3 cells were added and allowed to adhere to the IL-6-stimulated HBME, OC and OB cells for 30 minutes with 5% CO₂ at 37°C. Subsequently, non-adherent PC3 cells were discarded by removing the medium from wells and washing twice with HBSS. Plates were read using a fluorescence reader at excitation 485 nm and emission 530nm.

2.5 CXCR5 and α_vβ₃ integrin clustering

PC3 cells cultured with RPMI containing no additions, 1.0 nM of pokeweed mitogen-A (PMA; binds poly N-acetyllactosamines and induces integrin clustering), or 100 ng/mL of CXCL13 along with or without 100ng/mL of pertusis toxin (PTx; inhibitor of Gαi proteins) or 4 nM of wortmannin (inhibitor of PI3K) for 10 minutes. Untreated cells were used as controls. Next, cells were washed with FACS buffer (1% BSA in PBS) and then incubated with 1.0 µg/mL of FITC-conjugated anti-α_vβ₃, PE-conjugated anti-CXCR5, or isotype control Abs (R&D Systems) for 40 minutes in slurry of ice. After staining, unbound Abs were removed by washing with 500 µL of ice-cold FACS buffer three times. Cells were transferred to slides by cytospin (Shandon) and CXCR5 expression and α_vβ₃ clustering were analyzed by fluorescent microscopy.

2.6 Statistics

Levels of chemokines, PSA, and cytokines were expressed as the mean ± SEM (standard error of mean) and compared using a two-tailed student's t-test. Correlation analyses were performed between PSA and CXCL13 as well as cytokines and CXCL13 or PSA using SAS version 9.1.3 statistical analysis software. Pierson’s correlation was used to determine the correlation between the Log₁₀ value of serum PSA, CXCL13, IL-1β, TNF-α, and IL-6. Significance was declared at an α level < 0.01. A Venn diagram for regression display of the total sum of squares (TSS) as a rectangular box was used to illustrate differences in the coefficient of determination (R²) of serum CXCL13 and PSA. Sums of squares of CXCL13 and PSA are depicted as circles.

3. Results

3.1 Serum CXCL13, PSA, and cytokine levels in subjects with prostatic disease or healthy donors

Serum CXCL13 was significantly (p < 0.001) higher in PCa patients compared to patients with HGPIN or BPH and NHD (Figure 1). As expected, total PSA levels in PCa patient serum was significantly higher (p < 0.001) than compared to HGPIN, BPH, and NHD subjects (Figure 1B). Interestingly, CXCL13 serum levels seemed to provide a better distinction among PCa, BPH, HGPIN, or NHD subjects than PSA (Figure 2A). All PCa patients had serum CXCL13 levels > 75 pg/mL, which clearly differentiated PCa from other cases (BPH and HGPIN) or NHDs. However, this distinction could not be made when PSA was used as a prostatic disease predictive factor. In fact, total PSA > 10 ng/mL was associated with BPH as well as PCa cases. In contrast, 100% of the subjects with PCa had CXCL13 > 75 pg/mL. Perhaps the Venn diagram representation of this data set best represents the potential utility of CXCL13 as a biomarker for PCa progression (Figure 2B). While serum PSA levels for NHDs do not extend to patients with prostatic disease, the levels observed in BPH cases overlap with those concentrations measured in patients with HGPIN and PCa. In contrast, there is a clear distinction between serum CXCL13 levels from NHDs and patients with prostatic disease. While a small degree of overlap exists between BPH and HGPIN CXCL13 levels, there is no overlap and significant differences between serum CXCL13 concentrations in PCa patients than compared to BPH or HGPIN cases.

Figure 1. PSA and CXCL13 levels in serum of PCa, HGPIN, BPH, and NHD subjects.

ELISA assays were performed to quantify CXCL13 (A) and total PSA (B) levels in serum from patients diagnosed with PCa (n = 15), HGPIN (n = 5), or BPH (n = 10) and normal healthy donors (NHD, n=10). ELISAs were capable of detecting > 5 pg/mL of CXCL13 or PSA. Open circles indicate individual serum CXCL13 or PSA levels and lines show median concentrations of each group. Asterisks (*) show significant differences (p < 0.01) between groups with prostatic disease and NHDs.

Figure 2. Correlation between PSA, CXCL13, and serum cytokines levels of PCa, HGPIN, BPH, and NHD subjects.

ELISA assays were performed to quantify CXCL13, PSA, IL-1β, TNF-α, and IL-6. (A) CXCL13 and PSA levels in PCa (prostate cancer), HGPIN (high grade prostatic interstitial neoplasia), BPH (benign prostatic hyperplasia) and NHD (normal health donors) serum samples were graphed to determine associations with prostatic disease. Red, blue, orange and cyan circles represent PCa, HGPIN, BPH and NHD subjects, respectively. (B) The Venn diagrams are presented as total sum of squares (TSS) of PSA (TSS = 4294838824) or CXCL13 (TSS = 90124). Sums of squares for CXCL13 and PSA are depicted as red, blue, orange, and cyan circles for PCa, HGPIN, BPH, and NHD cases. (C) IL-1β, TNF-1α and IL-6 serum levels in PCa, HGPIN, and BPH subjects are expressed as means ± SEM and asterisks (*) indicate significant (p < 0.01) differences in cytokine expression between PCa and HGPIN or BPH. (D) Correlation matrix of PSA, CXCL13 IL-1β, TNF-α, and IL-6 serum levels from all subjects; values in bold indicate significant correlations.

Serum IL-1β, TNF-α, and IL-6 levels were measured to better determine the factors driving CXCL13 expression. While serum IL-1β and TNF-α levels were modestly higher in subjects with PCa, they were not significantly higher than those found in HGPIN or BPH cases (Figure 2C). On the other hand, significantly higher serum IL-6 levels were observed in PCa patients compared to other subjects. PSA levels did not significantly correlate with IL-1β, TNF-α or IL-6 (Figure 2C). However, significant correlations were found between IL-6 and CXCL13 (R² = 0.6378, p < 0.0003) as well as PSA and CXCL13 (R² = 0.9337, p < 0.0001). These results suggest that serum IL-6 level is associated with serum CXCL13, while serum CXCL13 and total PSA correlates with the extent of prostatic disease and in particular – PCa.

3.2 CXCL13 produced by bone stromal cells and CXCR5-dependent PC3 cell invasion

PCa cells most frequently metastasize to bone. We further investigated the ability of HBME to produce CXCL13 following IL-6 stimulation, since this cytokine is produced by bone marrow stromal cells during advanced PCa [28; 29; 30]. Untreated HBME, OB and OC cells produced CXCL13; however, only HBME and OBs treated with IL-6 significantly increased the expression of CXCL13 (Figure 3A). The CXCL13 produced by HBME and OB cells after IL-6 treatment was capable of inducing the invasion of PC3 cells through Matrigel™, which was inhibited by CXCR5 blockade (Figure 3B and 3C). These results suggest that IL-6 stimulates HBME and OB cells to secrete CXCL13 to promote CXCR5 positive cells to invade extracellular matrix (ECM) components. However, CXCR5 blockade did not completely abrogate PCa cell invasion suggesting that other chemotactic factors were also produced by HBME and OB cells, following IL-6 stimulation.

Figure 3. CXCL13 secretion by stimulated endothelial, OC, and OB cells and CXCR5-dependent PC3 cell invasion.

(A) Human bone marrow endothelial (HBME), osteoclast (OC) and osteoblast (OB) cells were cultured and treated with IL-6 (17.5 pg/mL). Conditioned media from untreated and treated cells were collected and CXCL13 was detected in supernatants by ELISA capable of detecting > 5 pg/mL of CXCL13. Asterisks (*) indicate significant (p < 0.01) differences in CXCL13 secretion between untreated and treated cells. (B and C) HBME or OB cells were grown in media alone (no additions, NA) or 17.5 pg/mL of IL-6 (metastatic PCa disease levels) overnight. Untreated or PC3 cells treated with 1 µg/mL of isotype control or anti-CXCR5 Abs for one hour and then allowed to invade matrigel chambers in response to CXCL13 secreted by HBME or OB cells in the bottom chamber. Asterisks (*) indicated significant differences (p < 0.01) between conditions.

3.3 CXCL13-mediated PC3 cell adhesion, αvβ3 clustering, and CXCR5 aggregation

PC3 cells adhered to HBME but not OB cells (Figure 4A). Treatment of HBME cells with 1.0 pg/mL of IL-6 (i.e., physiologically normal serum levels) did not significantly affect PC3 cell adhesion to endothelial cells. However, stimulation of HBME cells with higher amounts of IL-6, associated with localized (1.5 pg/mL) or metastatic (17.5 pg/mL) PCa, significantly increased PCa cell adherence to these cells. The increase in adherence to HBME cells was significantly reduced by anti-CXCR5 Ab treatment. Chemokine signaling has been shown to induce integrin clustering to support migration and/or adhesion [31]. As with PMA positive controls, CXCL13 treatment of PC3 cells induced rapid aggregation of CXCR5 and α_vβ₃ (Figure 4B). This clustering effect was not due to any increase in CXCR5 or α_vβ₃ expression, since the mean fluorescence of stained cells did not change after 10 minutes of CXCL13. This aggregation was partially mediated by PI3K and Gα_i protein subunit(s) signal transduction, since wortmannin and PTx, respectively, did not completely abolish this aggregation. Thus, PI3K- and G_i protein-dependent and -independent pathways support integrin clustering for adhesion and/or invasion following CXCR5 activation.

Figure 4. CXCR5-dependent PC3 cell adhesion and integrin clustering.

(A) HBME cells or osteoblasts were grown in complete media alone (no additions, NA) or with 1 pg/mL of IL-6 (i.e., normal serum level), 1.5 pg/mL of IL-6 (i.e., localized PCa serum level), or 17.5 pg/mL of IL-6 (metastatic PCa serum levels) overnight. Untreated or PC3 cells were treated with 1 µg/mL of isotype control or anti-CXCR5 Abs for one hour and allowed to adher to HBME or OB cells for 30 minutes. Asterisks (*) indicate significant differences (p < 0.01) between control and anti-CXCR5 Ab-treated groups. (B) PC3 cells received no additions (NA) or were treated with 1 nM of pokeweed mitogen A (PMA), 100 ng/mL of CXCL13, 100 ng/mL of pertussis toxin (PTX), and/or 5 nM of wortmannin for 5 or 10 minutes. After staining with PE-conjugated anti-CXCR5 or FITC-conjugated anti-α_vβ₃ antibodies, cells were fixed using 2% formaldehyde and added to slides by cytospin for examination by fluorescence microscopy at a 60× magnification.

4. Discussion

Despite recent advances in the treatment and diagnosis of cancer, PCa remains the second leading cause of cancer-related deaths among men in US. Detecting PCa at earlier curable stages has been facilitated by the measurement of the PSA in serum [32]. The American Urological Society, the American Cancer Society, and the American College of Physicians recommend that PCa screening to be discussed with all men over 50 years of age, as well as men over the age of 40 with additional risk factors such as African American race or a family history of PCa. Traditionally, PSA values > 4 ng/mL trigger a suspicion for PCa [33]. Men with serum PSA > 10 ng/mL have an increased risk of non-organ-confined PCa [5; 34; 35]. However, only 30% of men with serum PSA levels between 4 and 10 are diagnosed with PCa [36] and up to 15% of patients with PSA values < 4 carry the diagnosis [5]. In some cases of poorly differentiated disease, PSA levels are not significantly elevated. It is clear that PSA is an imperfect marker for PCa and that additional diagnostic and prognostic tools are needed. For the first time we show that CXCL13 is expressed at significantly higher levels in PCa patient sera, than compared to HGPIN, BPH, or NHD subjects. Hence, detection of serum CXCL13 might be used to refine the PSA test to better determine the risk or prognosis of PCa.

Metastasis to bone remains a major clinical concern and the main cause of PCa morbidity. The reasons for preferential establishment and growth of disseminated PCa cells to bone marrow are incompletely understood and require further investigation. Serum IL-1β, TNF-α and IL-6 levels are significantly elevated in patients with advance PCa [24; 25; 26; 27]. It is plausible that these inflammatory cytokines provide signals that subsequently lead to the production of additional inflammatory or growth factors for the host to “defend” against these danger signals. In this regard, we show that HBME and OB cells produce CXCL13, when treated with IL-6. Therefore, our data suggest IL-6, which is elevated during PCa, increases the production of CXCL13 by HBME and OB cells in bone marrow. Taken together, this supports the notion that CXCR5-expressing PCa cells migrate, invade and establish metastatic lesions in bone.

Bone marrow stromal cells as well as OB and OC cells are responsible for the bone remodeling process and homeostasis [37]. This process is significantly influenced by cytokines [38]. Our data show that CXCL13 is secreted by HBME and OB cells following IL-6 stimulation; hence, it is plausible that the CXCL13 secreted by these stromal cells are partially responsible for CXCR5⁺ PCa cell metastasis to bone. CXCL13 has also been shown to induce b-N-acetylhexosaminidase (Hex) release and alkaline phosphatase activity of OB cells [39]. Hex is an enzyme used to evaluate the standard response of leukocytes after chemokine stimulation, but it is also known to play a role in endochondral ossification and bone remodeling since it degrades glycosaminoglycans [40; 41]. Thus, CXCL13 or other chemokines that activate tumor-associated cells may also support bone turnover for the formation of osteoblastic lesions during advanced PCa.

Cytoskeleton re-modeling, adhesion, and de-adhesion are not only required for cell motility, but are also linked to proliferation and pro-survival pathways. One of the hallmarks of PCa cells is the lack of adhesion-dependent growth control as well as motility due to altered integrin signaling. The migration of PCa cells to the metastatic sites correlates with changes in integrin expression. PCa cells isolated from bone express α_vβ₃, which has a high affinity for fibronectin and mediates adhesion to vitronectin and osteopontin, which are major components in bone ECM [42; 43; 44]. Our data demonstrate co-localization of α_vβ₃ with CXCR5 and clustering of this integrin complex after CXCL13 treatment. This suggests that the CXCR5-CXCL13 axis is not only involved in PCa cell migration and invasion, but also adhesion. These studies reveal important cellular and molecular mechanisms of CXCR5- and CXCL13-mediated invasion and adhesion as well as support the rationale for future studies to target this axis to treat PCa. Moreover, CXCL13 detection in serum might be useful in enhancing PSA testing to improve the diagnosis of PCa and its separation from BPH or HGPIN.